Final Presentation by Visualization team

1
Final Presentation by Visualization team

• Team members
• Haibo Liu
• Robertas Baronas
• Jonathan Krentel
• Zaixia Zhang

2
Introduction, Requirements and Website

• By Haibo Liu

3
Final Presentation

• Background
• Requirements
• Input Files for First Iteration
• Web site

4
Background

• Radviz
• Visualization--n dimensions?2 dimensions
• Our project--data table?graph
• Terms
• target/anchor/
• division/
• attribute/numeric attribute/non-numeric
  attribute/
• data point/graph point
• normalization/

5
Requirements(1)

• Input file formatscomma-separated
  txt/excel/oracle/access/xml
• Target attribute
• Anchorsonly numeric attributes/25
• Data100/missing value/dirty value/dirty
  attribute
• Web download version
• Early prototypefinished!

6
Requirements(2)

• User--opens files
• Usersselects target/anchors
• Our systemoffers some useful information and
  suggestions
• Usersplays on the graph
• Userssave a session/graph or print

7
Input Files for 1st Iteration

• Names File
• Data File

8
Names File

• Description of data file
• Format
• --non-numeric name, nominal
• example country, nominal
• --numeric
• 1. name, numeric
• example number of cylinders, numeric
• 2. name, numeric, unit
• example price, numeric, dollar

9
Why?

• Sometimes, hard for our system to tell a value is
  numeric or non-numeric.
• Example telephone number, or bus number.
• So user has to tell us before using our system.

10
Web Site

• www.cs.umb.edu/visualization

11
Testing later

• Later, I am going to show the testings on our
  system, mainly about FileProcessor.

12
System Design, Architecture, Implementation

• By Jonathan Krentel

13
Design Concerns

• Adaptability to unforeseen new view needs
• Easy maintenance of view consistency
• Restrained but flexible exposure of internally
  held data
• Hot spot performance plotting data points
• Data structure divorce
• Avoidance of data redundancy
• Flexibility to potential new data sources

14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
(No Transcript)
18
Design Decisions

• Internally held data exposed via Focus
• User selections held, managed in Model
• Public getters, package setters and constructors
• Publicly available Division objects
• Objectification of system components
• Row based internal data model
• Heavy referencing, light construction
• Abstraction of data processing

19
Scenarios

• Please see handout

20
Open Concerns

• Data structure marriage
• Attribute data management
• DivisionSet exception handling
• GroupSet home
• Performance hot spots
• Dirty data handling
• Object-verb command structure
• Graph Architecture Outline

21
Schedule, Features, Databases Testing

• By Zaixia Zhang

22
Schedule

• Old
• 10/29/03 11/24/03
• 26 days
• 11/25/03 12/30/03
• 35 days (holiday in)
• 01/01/04 02/15/04
• 45 days
• 02/16/04 04/30/04
• 75 days

• Revised
• 10/29/03 12/16/03
• 48 days (holiday in)
• 12/17/03 01/25/03
• 40 days (holiday in)
• 01/26/04 03/15/04
• 50 days
• 03/16/04 04/30/04
• 45 days

23
Schedule

• Old

• Revised

24
Features (1)

• Deal with missing value, dirty value
• Handle equal values between minimum and maximum
  for given attribute
• Report statistic information for users data and
  give suggestions to users for selecting anchors
• Display outer big circle
• Compute and display anchors, data points
• Set different colors to data points for target
  with nonnumeric data type and numeric data type
• User can select target,anchors, add anchors

25
Features (2)

• Table view display
• User can input value range in each division after
  selecting target, display target range with color
• User can move anchor points by using mouse
• Anchor locations are recomputed and anchors are
  reordered
• The locations of data points are recomputed and
  data points are re-plotted
• F. Add, remove anchors by mouse dragging

26
Features (3)

• Display menu system, like file, tool, help with
  their submenus
• Save, new, open a section
• Display the original data table when user
  requests
• Compute and display the t-test value for every
  two groups, the correlation values for every two
  numeric attributes
• User can set the base value for correlation
• Display the data points information and anchor
  information when user do right click on the data
  points or anchors

27
Features (4)

• User can select an area in the graph, do zoom in
• Do histogram to show the range of target
  attribute
• Sort attributes in the table with original values
  
• Blinking/jingle data points
• Access different data source
• Design a nice web page and put our system on,
  make it downloadable

28
UCI Repository of Databases

• Contains wide set of different databases
• Tree structured, table structured
• What are we interested in?
• Table structured data sets
• Mixed with numeric/non-numeric attributes
• The number of attributes large enough
• Enough number of instances

29
Data Source Collection

• Data set collection
• Full or part of real data set from UCI repository
• Data type transfer
• Continuous, integer ? numeric
• Boolean, nominal ? nonnumeric
• Unit ? add one if it has, otherwise skip

30
Testing

• Test strategy
• Test by use cases
• Unit test by developer
• Integration test by two developer
• System test by specified tester
• Test cases (7)
• Test equal values, missing/dirty data, accuracy,
  special cases, capacity, color distribution

31
Testing

• Bug reports and fixes
• Report by bug report form (one bug per form)
• Fixed by developer or tester
• Test assumption
• Input data is comma separated, its format
  following our specification
• First user select target, then select anchors
• Can not handle Boolean type attribute

32
Test Case 4 Accuracy Test

• Plotting algorithm (RADVIZ approach)
• Data segment

33
Test Case 4 Accuracy Test

• By excel

• By PlottingAlgorithm

34
Test Case 5 Special Case

• No anchor is selected
• One anchor is selected
• 2 anchors are selected
• 25 anchor are selected

35
Test Case 6 System capacity

• Data source
• case6/housing-names.txt (14 attributes, one of
  them is nominal)
• case6/housing-data.txt (number of instances 506)
• Database housing
• No missing values.

36
Test Case 7 Color Distribution

• Data source
• case6/bignames.txt (9 attributes, 4 of them are
  nominal, car name is unique)
• case6/bigdata.txt (number of instances 392)
• Database auto-mpg
• No missing values.

37
Thanks!

• More questions?